Horton

Re: hung backends stuck in spinlock heavy endless loop

Peter Geoghegan <pg@heroku.com>

View thread View on postgr.es
From: Peter Geoghegan <pg@heroku.com>
To: Merlin Moncure <mmoncure@gmail.com>
Cc: Andres Freund <andres@2ndquadrant.com>, Tom Lane <tgl@sss.pgh.pa.us>, PostgreSQL-development <pgsql-hackers@postgresql.org>, Heikki Linnakangas <hlinnakangas@vmware.com>
Date: 2015-01-14T17:42:24Z
Lists: pgsql-hackers 
On Wed, Jan 14, 2015 at 7:22 AM, Merlin Moncure <mmoncure@gmail.com> wrote:
> I'll try to pull commits that Peter suggested and see if that helps
> (I'm getting ready to bring the database down).  I can send the code
> off-list if you guys think it'd help.

Thanks for the code!

I think it would be interesting to see if the bug reproduces without
the page split commit (40dae7ec537c5619fc93ad602c62f37be786d161) first
(which was applied afterwards). Then, without the page deletion commit
(efada2b8e920adfdf7418862e939925d2acd1b89). But without the page split
commit seems most interesting at the moment. It would be easier to
diagnose the issue if the bug still reproduces without that commit -
that way, we can be reasonably confident that there are no confounding
factors from the new page split code.

There were some bugfixes to those two during the 9.4 beta cycle, too.
It might be a bit tricky to generate a 9.4 that lacks one or both of
those commits.

Bugfix commits in reverse chronological order:

c73669c0e0168923e3f9e787beec980f55af2bd8 (deletion)

c91a9b5a285e20e54cf90f3660ce51ce3a5c2ef4 (incomplete split)

4fafc4ecd9e4d224d92c4a8549c5646860787a5d (deletion)

4a5d55ec2b711e13438a32d119a809a22ced410b (incomplete split)

77fe2b6d795f3f4ed282c9c980920e128a57624e (deletion)

7d98054f0dd115f57ad0ec1f424a66c13459013b (deletion)

954523cdfe229f1cb99a43a19e291a557ae2822d (incomplete split)

I think that's all of them (apart from the original commits, of course).

-- 
Peter Geoghegan

Commits

  1. Improve hash_create's API for selecting simple-binary-key hash functions.

  2. Fix WAL-logging of B-tree "unlink halfdead page" operation.

  3. Fix backup-block numbering in redo of b-tree split.

  4. Cleanup of new b-tree page deletion code.

  5. Fix bug in the new B-tree incomplete-split code.

  6. Fix bug in new B-tree page deletion code.

  7. Fix rm_desc routine of b-tree page delete records.

  8. Make the handling of interrupted B-tree page splits more robust.

  9. Fix race condition in B-tree page deletion.